Watermark-based error concealment algorithms for low bit rate video communications

In this work, a novel set of robust watermark-based error concealment (WEC) algorithms are proposed. Watermarking is used to introduce redundancy to the transmitted data with little or no increase in its bit rate during transmission. The proposed algorithms involve generating a low resolution version of a video frame and seamlessly embedding it as a watermark in the frame itself during encoding. At the receiver, the watermark is extracted from the reconstructed frame and the lost information is recovered using the extracted watermark signal, thus enhancing its perceptual quality. Three DCT-based spread spectrum watermark embedding techniques are presented in this work. The first technique uses a multiplicative Gaussian pseudo-noise with a pre-defined spreading gain and fixed chip rate. The second one is its adaptively scaled version and the third technique uses informed watermarking. Two versions of the low resolution reference, a halftoned reference and a DPCM encoded reference, are considered here. Spatial, temporal, and spatio-temporal implementations of WEC are proposed for video. A reference watermark of either the intra-coded frame or the subsequent intercoded frame is embedded in the current frame for mitigation of error propagation. In the case of spatio-temporal implementation, a low resolution gray-scale reference is bit-plane embedded in a volume data set. These implementation schemes not only enhance the end-user perceptual video quality, but also increase the embedding capacity. Both qualitative and quantitative analysis of the WEC algorithms along with a comparison between the full-frame and block-based embedding techniques are presented. A psychophysical experiment is performed to obtain the subjective quality evaluation of the proposed techniques, the comparison between the perceptual quality of intracoded reference embedding and inter-coded reference embedding, and to verify the codec independency of the WEC algorithms. Experimental results show that the proposed WEC techniques outperform other current error concealment algorithms, especially at higher transmission losses. In video implementations, the inter-coded frame embedding resulted in nearly a constant perceptual quality performance when compared to the intra-coded frame embedding scheme. The psychophysical experimental analysis has confirmed this observation.